Performance evaluation of different seasonal adjustment techniques

Abstract

The central Bank of Bangladesh publishes financial data on weekly, quarterly, monthly, yearly basis. Among the published data, GDP, production, export, import, balance payment, wage earners remittance, Stock etc. are time-based which often subject to seasonal variation. But like most of the developed country's central bank such as USA, Australia, Singapore, UK, French etc. Bangladesh Bank does not publish seasonal adjusted data. These published time series are used by economists, decision-makers, and consumers to inform decisions. They look for key characteristics of economic series like trend, turning points, and coherence with other economic data. Seasonal shifts might occasionally make it challenging to see these features. Therefore, the research aimed to explore the best seasonal adjustment technique for different financial data published by Bangladesh Bank. Here, non-filter-based seasonal adjustment technique: Ratio to moving average, Ratio to trend, and SARIMA and filter-based technique: X-11 and X-13-ARIMA-SEATS was applied on monthly import and export data published by Bangladesh Bank. The seasonal adjustment method that produces fewer forecasting errors (MAPE, MASE, MAE, and RMSE) were decided as best technique for seasonal adjustment and X-13-ARIMA-SEATS method founds the appropriate seasonal adjustment method for our considered data. Therefore, X-13-ARIMA-SEATS method can be recommended to forecast monthly export and import of Bangladesh published by Bangladesh Bank.